IGI combiner utility with source code

[mgix]

I have hit a scalability glass ceiling with the current Indigo network
rendering model for (sadly, master explodes past ~600 slaves, and
even when it doesn't, it becomes the bottleneck).

This has reduced me to have each "slave" dump IGIs to local disk,
collect them via scp and "manually" accumulate them into one final IGI.

To perform the final accumulation step, I wrote a litte tool that I thought
others might find useful.

There already is a tool that does this somewhere else in the forums, but
since I don't use windows and since the other tool that does has no source
code attached, I went a rewrote one myself.

It should work with recent IGIs (version 4, with layers).

Source code is here:

http://www.mgix.com/snippets/igiacc.cpp

Compile with:

Code: Select all

g++ -g0 -O3 -fomit-frame-pointer -o igiacc igiacc.cpp

Windows binary is here:

http://www.mgix.com/snippets/igiacc.exe.bz2

Use as follows:

Code: Select all

igiacc *.igi

Enjoy,

- Mgix

[PureSpider]

Does this compile on windows?

[mgix]

Does this compile on windows?

Now it does.

[CTZn]

Your contribution is much apreciated mgix... necessity is the mother of all innovations

You can find the equivalent for windows, from dougal2:

http://www.indigorenderer.com/joomla/fo ... t=igimerge

[SATtva_]

mgix, I have trouble compiling it under amd64 with both GCC 3.4:

Code: Select all

/tmp/ccoQbIYZ.o:(.gnu.linkonce.d.DW.ref.__gxx_personality_v0+0x0): undefined reference to `__gxx_personality_v0'
collect2: ld returned 1 exit status

...and 4.1/4.2:

Code: Select all

/tmp/ccxKzTLI.o: In function `isBigEndian()':
igiacc.cpp:(.text+0x26): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x42): undefined reference to `__cxa_guard_release'
/tmp/ccxKzTLI.o: In function `main':
igiacc.cpp:(.text+0x406): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x42b): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x44b): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x45a): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x47a): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x4d4): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x54a): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x56a): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x579): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x599): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x5b7): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x5d7): undefined reference to `__cxa_guard_release'
igiacc.cpp:(.text+0x683): undefined reference to `__cxa_guard_acquire'
igiacc.cpp:(.text+0x69f): undefined reference to `__cxa_guard_release'
/tmp/ccxKzTLI.o:(.eh_frame+0x12): undefined reference to `__gxx_personality_v0'
collect2: ld returned 1 exit status

Any assumptions or prerequisites I failed to met?

[mgix]

mgix, I have trouble compiling it under amd64
Any assumptions or prerequisites I failed to met?

No, I just wrote my post too hastily.

The correct command line to compile should read 'g++' instead of 'gcc'.

I'll go correct my post now.

Cheers

[SATtva_]

Yeah, this makes much difference. =)

[OnoSendai]

Hi Mgix,
I've updated the IGI spec here:
http://www.indigorenderer.com/joomla/fo ... sc&start=0

[mgix]

Hi Mgix,
I've updated the IGI spec here:
http://www.indigorenderer.com/joomla/fo ... sc&start=0

Hey Ono,

Thanks for taking the time to update the specs, but ...
if you happen to have time, a little more is needed

Because the compiler is free to align and pad things any way it
chooses, only giving a C++ class layout isn't enough: it makes
your spec compiler dependent.

As a practical example of what I mean, gcc and VC++ currently
have different ideas of what sizeof(IndigoImageHeader) version 4 is.

I copied the class code verbatim from your spec into a small c++
program, and ...

With GCC: sizeof(IndigoImageHeader) is 5028
With VC++: sizeof(IndigoImageHeader) is 5032

Which is the correct answer ?
Does the image data start at offset 5032 or at offset 5028 ?

It could potentially be inferred from the overall size of the file
and the spec'd size of the pixel data, but I don't think that'd be
a very nice thing to have to rely on.

Also, if you do rely on sizeof(IndigoImageHeader) inside indigo's code
when writing out IGIs, your format will not be stable across platforms,
compilers or even different releases of the same compiler ...

[mgix]

Update: to answer the plea of the pour souls still stuck with windows,
I have ported igiacc. I've also added a link to a binary in the original
post.

[OnoSendai]

Hi Mgix,
I've updated the IGI spec here:
http://www.indigorenderer.com/joomla/fo ... sc&start=0

Hey Ono,

Thanks for taking the time to update the specs, but ...
if you happen to have time, a little more is needed

You are of course quite right mgix. I've updated the IGI spec to add another 4 bytes to the padding, to reflect the size of the structure that VS was producing. I've also added a more explicit description of the padding, or lack thereof, that should be expected on disk.
Sorry about that!